Anthropic filing: existing customers' Claude spend grew ninefold in a year
Anthropic's IPO filing says existing customers spend nine times what they did a year ago. Most seats cost under $75 a month; the heaviest cost 20 times that.
Businesses that were already paying Anthropic in the third quarter of 2025 spent about nine times as much with it in the third quarter of 2026. That is the "Net Revenue Retention of approximately 900% as of September 30, 2026" in the [amended IPO filing](https://www.sec.gov/Archives/edgar/data/2130357/000162828026074338/anthropicpbc-sx1ax3.htm) Anthropic lodged with the SEC on 17 November 2026. The measure divides those customers' revenue in the latest quarter by their revenue a year before, counts those that cut back or left, and leaves out people on the fixed-price Pro, Max and Ultra plans.
For whoever owns the AI budget, the filing is the seller's own account of what happens to a bill charged by use, not by seat, once staff have agents: software that carries out multi-step work on its own.
Most seats are cheap; a few are not
The filing says: "in the three months ended September 30, 2026, Claude Enterprise users who did not use Claude Code spent less than $75 per month on average, while the top 5% of Claude Enterprise users who used Claude Code spent approximately 20 times that amount." Claude Code is Anthropic's coding agent. Twenty times $75 is $1,500, so the heaviest users cost up to about $1,500 a month each. The filing gives no median, and no equivalent figure for Claude Cowork, the agent it sells to non-engineers.
Customers paying $100,000 or more over twelve months rose from more than 950 at 30 September 2025 to more than 11,900 a year later. More than 1,900 paid $1 million or more, and about 250 paid $10 million or more.
Anthropic's own usage passed its cash payroll
Anthropic had about 5,300 employees at 30 September 2026, a majority in research, engineering and product. It writes: "The value of our own internal Claude usage at list prices has grown by over 250 times from January 2026 to September 2026, and in April 2026, this monthly value first exceeded our monthly personnel expense, excluding stock-based compensation."
That is usage valued at the published price, not what it cost Anthropic to run, and the filing gives no dollar figure and no split by team. Much of it is engineering: more than 90% of contributions to its codebase in September 2026 were written by Claude, and engineers now "direct and review its work rather than write code line by line". For its sales and operations teams the filing gives no numbers, only that Claude lets "a comparatively small team" do the work of a larger one. Headcount still grew more than five times between December 2024 and September 2026.
Reading this at work? Get one story like it explained every morning.
A five-minute read every morning at 7am ET: one story explained, two in brief, what else happened, one number. Unsubscribe in one click.
Prices fell and bills still rose
Anthropic says its Fable 5.1 model, released in September 2026, cut the cost of typical workloads by about 25% against Fable 5, and Opus 5.5 cut customers' cost per task by about 40% against Opus 5. Price cuts and contractual discounts took about $600 million off revenue growth in the third quarter; revenue for the quarter was still $17.0 billion. "Lower costs per task drive greater usage," the filing says.
The catch
The filing says Net Revenue Retention and its customer counts "have not been independently verified by third parties". The retention figure is weighted by revenue, so the largest spenders dominate it; it is not what a typical customer's bill did.
Two terms matter to a buyer. "A growing number of enterprise customers agree to non-cancelable annual or multi-year contracts with minimum spend commitments in exchange for a committed price list." And Anthropic says it has "changed, and may in the future change, usage limits, rate limits, or model and product capabilities".
What to do before the next AI renewal
- Finance: ask the vendor, or the admin console, for last quarter's spend per user. Budget the top 5% on a separate line from everyone else.
- IT: before switching on agent features, ask how spend is capped per user or per team, set the cap, and name who can raise it.
- Whoever signs: before agreeing a minimum spend, get in writing what happens to unused commitment and to your price list if usage limits change mid-term.
Take this to work
Forward this
Anthropic's SEC filing of 17 November 2026 says customers already paying a year ago now spend about nine times as much, and that the top 5% of Claude Enterprise users who use its coding agent spend about 20 times the under-$75 a month of users who do not. Our AI budget should assume usage-based spend grows after agents are switched on, with a cap per user and a separate line for the heaviest users.
Questions to ask
- What did each of our users spend last quarter, and what share of the bill came from the top 5%?
- Can spend be capped per user or per team, and what does a user see when they reach the cap?
- If we commit to a minimum spend, what happens to the unused part at the end of the term?
- If usage limits or rate limits change mid-contract, does our committed price list still hold?
- Which of our teams would use agents heavily, and who reviews their output?
The arithmetic
(users who do not use agents × their average monthly spend) + (heaviest users × their average monthly spend) = monthly bill
The filing gives under $75 a month for Claude Enterprise users who did not use Claude Code, and about 20 times that (20 × $75 = $1,500, so up to about $1,500) for the top 5% of users who did. It does not give the spend of Claude Code users outside the top 5%, the share of users who use Claude Code, or any figure by team, so a full bill cannot be built from it.
Questions this article answers
What is Anthropic's net revenue retention?
About 900% as of 30 September 2026, according to its amended Form S-1. Customers with revenue in the third quarter of 2025 generated about nine times as much in the third quarter of 2026; Pro, Max and Ultra subscribers are excluded.
How much does a Claude Enterprise user cost per month?
Anthropic's filing says Claude Enterprise users who did not use Claude Code spent less than $75 a month on average in the third quarter of 2026. The top 5% of users who did use Claude Code spent about 20 times that amount.
How much does Anthropic spend on its own use of Claude?
It gives no dollar figure. The filing says the value of its internal Claude usage at list prices grew more than 250 times between January and September 2026 and passed its monthly personnel expense, excluding stock-based compensation, in April 2026.
Do Anthropic enterprise contracts have minimum spend commitments?
The filing says a growing number of enterprise customers agree to non-cancelable annual or multi-year contracts with minimum spend commitments in exchange for a committed price list.
LegalZoom cut 13% of staff to go 'AI-native'; its profit forecast fell $10M
LegalZoom's filings put a 13% staff cut at $6.0M in severance. Three months on, dearer customer acquisition took $10M off its profit forecast.
Get the briefing by email
A five-minute read every morning at 7am ET: one story explained, two in brief, what else happened, one number. Unsubscribe in one click.